A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 4, 16:15 UTC
[Test] Increase qwen2_vl num_logprobs to fix torch 2.12 update (#40818)
Signed-off-by: Angela Yi <angelayi@meta.com>
Angela Yi committed
Apr 24, 2026 at 17:59 UTC
bc2ae5a3d6b59690b6a3312f0ed63842e8bc600b
1 file changed
+1
tests/models/multimodal/generation/test_common.py
+1
@@ -921,6 +921,7 @@ VLM_TEST_SETTINGS = {
921
multi_image_prompt="Picture 1: <vlm_image>\nPicture 2: <vlm_image>\nDescribe these two images with one paragraph respectively.", # noqa: E501
922
max_model_len=4096,
923
max_num_seqs=2,
924
+ num_logprobs=10,
925
auto_cls=AutoModelForImageTextToText,
926
vllm_output_post_proc=model_utils.qwen2_vllm_to_hf_output,
927
image_size_factors=[(0.25,), (0.25, 0.25, 0.25), (0.25, 0.2, 0.15)],